A VLSI for deskewing and fault tolerance in LVDS links

• Published in: IEEE transactions on nuclear science Vol. 53; no. 3; pp. 801 - 809
• Main Authors: Torralba, G., Angelov, V., Gonzalez, V., Lindenstruth, V., Sanchis, E.
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.06.2006; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Bandwidth; Cables; CMOS logic circuits; CMOS process; Communication cables; Copper; Data transmission; Delay units; Devices; Dynamical systems; Farms; Fault tolerance; Frequency; granularity; Links; Logic; Logic devices; low-voltage differential signaling (LVDS); parallel links; skew compensation; Switches; Very large scale integration
• ISSN: 0018-9499; 1558-1578
• DOI: 10.1109/TNS.2006.874799

The need for high input/output (I/O) bandwidth has led to the use of point-to-point parallel links. At the same time, the low-voltage differential signaling (LVDS) technology has become very popular as a standard mostly due to its high performance and reliability in conjunction with low price. However, one limitation of the LVDS transmission is the strong dependence of the performance on the transmission distance since the media is a copper cable and data are transmitted at high frequency. The skew of the transmitted data is extremely important because it directly affects the sample window available to the receiver logic. It either forces to use high quality cables in order to minimize its effects or to reduce the maximum transmission distance. The device presented in this paper is a mixed digital/analog design implemented in a 0.35 /spl mu/m CMOS process for compensating the skew which affects parallel data transmissions and for providing fault tolerance in large scale systems, for instance used in trigger farms for high-energy physics experiments. The SWItch for Fault Tolerance chip (SWIFT) compensates dynamically skews of LVDS signals up to 250 MHz in steps of 100 ps and adds fault tolerance to a farm of PCs by allowing the bypassing of a failing compute node to which is attached.